5.3.

Boiling Heat Transfer

The prediction of the heat transfer coefficient to
boiling liquids is subject to large errors due to the
inability to specify, manufacture, and maintain the
nucleation characteristics of surfaces, as was
illustrated by Figures 5.2 and 5.5. However, the
boiling heat transfer coefficient is only part of the
overall coefficient and the effect of these large
errors is reduced in the overall design but must be
considered in selection of safety factors or the
selection of operating parameters; e.g., using only
a fraction of the available steam pressure. Boiling
heat transfer has been studied extensively and
good summaries of these researches are found in
(9, 10, 21, 22).

The design procedures and equations are
different for boiling outside of tubes or pool boiling
and for boiling inside of tubes. A single tube
boiling heat flux curve is a basic starting point and
the development of this curve is as follows. A
natural convection coefficient equation is used to
generate the natural convection heat flux curve by

q
nc
= h
nc
T (5.6)

and plotted in Figure 5.17 as line OA. Then a
nucleate boiling heat flux curve is calculated and
plotted as line AB. The intersection of these lines
is point A. A calculation of the maximum flux
determines point B on the nucleate curve. Usually
these two curves are sufficient but if the critical T
at point B can be exceeded, then the minimum flux and the corresponding temperature difference is
calculated and establishes point C and the film boiling flux curve CD drawn. No predictive equations exist
for the intermediate region (BC) but a straight line is drawn between these points. What we now have is
an approximation of the curve in Figure 5.1 by a series of straight lines. The real curve (as in Figure 5.1)
has a smooth transition between these straight lines but in design these transition curves are ignored and
results in a slight conservatism.

5.3.1. Pool Boiling - Single Tube

(a) Natural Convection

When the wall temperatures are too low to initiate nucleate boiling the heat transfer coefficient is based
on the liquid natural convection coefficient where the T is based on the difference between the wall and
the liquid temperatures. For horizontal tubes the following equation is used

254

25 .
2
2 3
53 . 0
l
l

=
k
c
T g d
k
d h p
o o nc

(5.7)

(b) Nucleate Boiling

Although several attempts (23, 24) of theoretical type equations utilizing the fluid properties have been
proposed, they are impractical because of the required physical property data (often unavailable for the
designer's problem), their complicated evaluation, and their inherent uncertainty due to the surface
conditions. A simpler approach, widely used by designers, is based on the work of Borishanski (25) who
utilized the law of corresponding states and, as modified by Mostinski (15) and Collier (26), is given as

h
nb
= A
*
q
0.7
F(P
r
) (5.8)

where A
*
is a constant evaluated at a reference reduced pressure of P
r
= 0.0294 and F(P
r
) is a function of
reduced pressure as shown in Figure 5.18.



A
*
= 0.00658
69 . 0
c
P (5.8a)

F(P
r
)
1
= 1.8 + 4 + 10 (5.9)
17 . 0
r
P
2 . 1
r
P
10
r
P

However, dropping the last two terms is a safe design (26) for pool boiling hence

255

F(P
r
)
2
= 1.8 (5.10)
17 . 0
r
P

However, for refrigerants (R11, R12, R113, R115, etc.), F(P
r
) should be evaluated as

F(P
r
)
3
= 0.7 + 2 P
r
[4 + 1/(1 P
r
)] (5.11)

and the values of A
*
for these refrigerants given in the square brackets in Table 5.1 be used in equation
5.8. (26)

It should be emphasized that the above equation was developed for single component liquids and that
the system pressure, tube surface condition, presence of non-condensable gases, hystersis of the boiling
curve, size and orientation of the surface, subcooling, wettability, and gravitation acceleration are some of
the variables that can affect the result.

(c) Critical or Maximum Heat Flux.

Cichelli and Bonilla (13) found that the maximum flux was a function of reduced pressure, P
r
= P/P
c
, and
that this curve also had a maximum at a reduced pressure of 0.3, Figure 5.4. Based on the assumption
that the maximum flux was limited by the hydrodynamic flow pattern at the surface, Zuber (24), developed
the following theoretical equation which seems to correlate the data and with a minor adjustment of the
theoretical constant, / 24, is

4 / 1
2
) (
18 . 0

=
v
v c
v cr
g g
q

l
(5.12)

This equation was for flat plates and some effect
of geometry is found which shows the constant
ranging from 0.12 to 0.2 (27) depending upon a
dimensionless parameter

2 / 1
) (

c
v
g
g
L
l


where L is radius or a length of plate. Curves are
given for several different shapes, sphere, plates,
and cylinders, Figure 5.19. In addition, other
factors such as liquid viscosity, subcooling, and
surface conditions can affect the values given by
equation 5.12. However, equation 5.12 is
generally used in commercial design as there are
other contributing factors in an actual exchanger
that influence the maximum.

(d) Minimum Film Boiling Flux.

The minimum heat flux, q
f
, for film boiling, point C in Figure 5.17, occurs when the minimum rate of vapor
formation to sustain a stable vapor film is reached. The Zuber theory for the minimum heat flux in film
boiling was improved by Berghmans (28) who considered second order perturbations and included the
256

effect of vapor film thickness into the analysis. From an analysis by Berenson (29) for flat plates and the
maximum flux equation 5.12, we find that

2 / 1
09 . 0
18 . 0

+
=
v
v
mf
cr
q
q

l
(5.13)

and that the corresponding T
min
is

3 / 1
2 / 1
3 / 2
min
) ( ) (
) (
127 . 0

=
v c
v
v
c
v
v
v
v
g g
g g
k
T

l l l
l
(5.14)

A straight line between points B and C in Figure 5.17 then is used for the so-called transition region.
Through point C a film boiling flux curve CD can be drawn using the heat flux calculated from the film
boiling coefficients.

(e) Film Boiling Heat Flux.

At high temperature differences a continuous vapor film covers the surface and analytic analysis has
followed an analogy to film condensation. For large horizontal plates

[ ]
4 / 1
3
2 /
) (
425 . 0

=


c
v
e
v v
f
L T
g k
h
l
(5.15)

where L
c
is the shortest unstable wave length for the Taylor instability given by

[ ]
2 / 1
2

=
v
c
c
g
g
L

l
(5.16)

and
e
is an effective latent heat including the superheat effect

e
= [1 + 0.4(c
pv
T/)] (5.17)

For tubes the equations are

4 / 1
' 3
) (
] / 069 . 59 . 0 [

+ =
c
v
e
v v
c
f
TL
g k
d L h

l
(5.18)

but here

257

e
= [1 + 0.34(c
pv
T/)]
2
(5.19)

These equations give the conduction heat
transfer but in addition at these temperature
levels radiation becomes important. Hence, the
total film coefficient as suggested by Bromley
(30) is

h
ft
= h
f
+ 0.75 h
r
(5.20)

where h
r
is the coefficient for radiation transfer
assuming the liquid is a black body and
radiation is between infinite parallel plates.

The flux is then calculated by equation 5.6.

5.3.2. Single Tube in Cross Flow

The effect of cross flow on the boiling
coefficients is shown in Figure 5.20. Basically
the forced convection coefficient may exceed
the nucleate boiling coefficient at low
temperature differences but at higher T the
nucleate boiling coefficient becomes dominant.
The simple rule is to use the highest coefficient.

The critical or maximum heat flux is also
affected by cross flow velocity but the
magnitude of this effect is a function of tube
size and seems to disappear when ap-
proaching industrially used dimensions as
shown in Figure 5.21.

In the film boiling region equation 5.18 will apply
for low cross flow velocities but for higher
velocities and for less than 45C (81F)
subcooling the following equation applies

2 / 1
7 . 2

=

sat o
v v
c
T d
k V
h

(5.21)

5.3.3. Boiling on Outside of Tubes in
a Bundle

In spite of the wide use of horizontal tube
bundles there is very little data and only
elementary suggestions for predicting its heat
transfer. The very early work of Abbot and
Comley (31) showed the coefficients for a
258

bundle and a single tube were essentially the
same. Other reports by Palen et al., (32)
showed that bundles may perform better than
single tubes due to the circulation induced
through the bundle. Although circulation models
are being developed, very little has been
published on these models. As shown in Collier
(26) and in Figure 5.22 the coefficients vary in a
haphazard fashion throughout the bundle but, in
general, increase from the bottom to top. Palen
(34) recommends using

h
b
= h
nbl
F
b
F
m
+ h
nc
(5.22)

where F
b
= 1.5, h
nc
= natural convection
coefficient, and h
nbl
= the single tube nucleate
coefficient. F
m
is the mixture correction. The 1.5
factor is a conservative approximation as it
could range up to 3 depending on a bundle
layout, size, and heat flux.

There is a maximum flux for a bundle that is different, and lower, than the single tube maximum flux.
Based on some plant experience Palen and Small (35) proposed a model assuming a vapor blanketing
effect. They developed a correction term,
b
, which is used to multiply the q
max
as calculated by equation
5.12 and corrected for mixture effects, eqn. 5.38. Their result can be simplified to

1 . 1
) / (
2 . 2
tp B o
tp
s
B
b
L D d
L K
A
L D
=

=

(5.23)

where K = 4.12 for square pitch
K = 3.56 for triangular pitch
(
b
)
min
= 0.1

This result is reported to be conservative by Palen et al. (32).

5.3.4. Boiling Inside Tubes

As shown in Figure 5.6 vaporization inside tubes involves a number of different flow regimes each of
which requires a different evaluation of the coefficient plus a local temperature difference which in turn
requires corresponding pressure drop calculations. Since most are of a natural circulation design, only
the available liquid head is known thus resulting in a trial and error series of calculations to determine the
feed rate per tube. The calculated procedure is, thus, too tedious for hand calculation and computers are
utilized. However, the computer programs are also complex and expensive to develop and, therefore,
become proprietary. We list below the various equations used for the design of vertical in tube vaporizers.
Although horizontal in-tube vaporization is also used, the heat transfer equations and methods are the
same but the flow pattern now includes a stratified two-layer region. The major difference between
horizontal and vertical in-tube vaporization is the definition of and the flow criteria used to define the limits
of each regime. The appropriate heat transfer equation is then used for each regime.

259

(a) Single Phase Liquid Region.

In a circulating vaporizer the temperature of the liquid entering the tube is below the local boiling point
due to the effect of the hydrostatic head on the saturation temperature. This liquid zone extends to the
point where the temperature has increased and the local pressure decreased such that the local
saturation point has been reached. Actually some further superheat is required to initiate nucleation. The
liquid zone heat transfer coefficients are calculated from

1 .
2
2 3
25 . 43 . 33 .
Pr
Pr
17 . 0

=
l
l l
l
l
l l

T g d
k
G d
k
d h
i
w
i i c
(5.24)

for L/d
i
> 50 and d
i
G/ < 2000.

For turbulent flow and d
i
G/ > 10,000 use

3 / 1 8 .
023 . 0

=
l
l
l l
k
c
G d
k
d h p
i i c

(5.25)

and interpolate between these two equations on a Re number basis for 2000 < d
i
G/ < 10,000.

(b) Boiling Region.

The boiling region can be further subdivided into a sub cooled boiling, saturated boiling, and two-phase
boiling regions with predictive equations for each (9,26). Another approach taken by Chen (36) is to
combine the saturated and two-phase regions into one, with an equation combining the convective and
nucleate boiling mechanisms

h
b
= s h
nb
+ h
cb
(5.26)

where h
b
= the boiling coefficient
h
nb
= the nucleate boiling coefficient
h
cb
= the convective coefficient
s = Chen suppression factor

The convective coefficient is a function of the Martinelli two-phase flow parameter, X
tt
, and the Chen
correlation using this factor is

ch tt
c
cb
F x f
h
h
= = ) ( (5.27)

73 . 0
213 . 0
1
35 . 2

+ =
tt
ch
x
F (5.28)

( )
11 . 0 57 . 0
1

=

v
v
x
x
tt
x

l
l
(5.29)
260


x = weight fraction of vapor
h
c
= liquid phase heat transfer coefficient based on the amount of liquid present, Equation 5.25.

The nucleate boiling coefficient, h
nb
, is determined as

h
nb
= h
nbl
F
m
(5.30)

where F
m
is a correction applied for mixtures (discussed later) and h
nbl
is the coefficient determined from
equation 5.8. The suppression factor, s, is determined as follows:

1. Calculate liquid phase
l l
/ Re G d
i
=

2. Calculate two-phase Re
tp
= (5.31)
25 . 1
Re
ch
F
l

3. Calculate s= 1/[.1 + 2.53(10
-6
) ]
17 . 1
Re
tp

Now equation 5.26 can be solved for h
b
.

The subcooled boiling coefficient can be obtained by again using equation 5.26 but with s = (T
b
/T
o
)
where T
b
is the temperature difference between the tube wall and the saturation temperature of the
liquid at the given local pressure and T
o
is the difference between the tube wall and subcooled bulk
temperature. Instead of the convective coefficient, h
nbl
, the liquid coefficient (eqn. 5.24 or 5.25) is used.
The nucleate coefficient, h
nbl
, is obtained from the transformed equation 5.8 as

h
nbl
= 5.43(10
-8
)(
c
P )
2.3
(T)
2.33
[F(P)]
3.33
(5.32)

and equation 5.32 changed to

h
nb
= h
nbl
F
m
(5.33)

where F
m
is from equation 5.38.


(c) Mist Flow.

In mist flow the small amount of remaining liquid is en trained as droplets and the tube wall is essentially
dry. The coefficient drops rapidly and approaches that of heat transfer to gas. In this regime sensible heat
is transferred to the gas which in turn transfers some of the heat to the droplets until they are completely
evaporated after which only sensible heat transfer to gas occurs. The main problem is the determination
of the vapor temperature, hence, temperature difference. Two extreme conditions are: (1) no heat is
transferred to the droplets hence the vapor temperature rises rapidly; and (2) heat is rapidly transferred to
the droplets until they disappear and during this evaporation phase the vapor is at saturation temperature.
Condition I is approached at low pressures and velocities and condition 2 at high pressure and velocities.
The actual case is somewhere between 1 and 2. Some attempts to develop empirical and theory based
equations are reported (26) but the range of data seem too limited. We would recommend to use an
equation like 5.25 based on gas properties and then make an engineering judgment guess of the fraction
of the sensible heat transferred to the gas that is used up as latent heat for the evaporation of drops. The
261

resulting effect on vapor temperature could be used to calculate an LMTD for the mist region and with the
calculated gas coefficient used to determine the heat flux.

The mist region can be determined from a Fair (37) map or from the simple equation derived from this
map

G
mm
= 1.8(10
6
) X
tt
lb/hr ft
2
(5.34)

where G
mm
is the maximum mass velocity before mist flow begins.

(d) Film Boiling.

This type of boiling should be avoided, if possible, due to control problems, possible fouling, and lack of
data on pressure drop calculations. But if the temperature difference is high enough over the entire tube
length, then the heat transfer coefficient can be calculated by the Glickstein and Whiteside (38)
correlation

Nu = 0.106 Re
0.64
Pr
0.4
(
b
/
v
)
0.5
(5.35)

where the bulk average density on a no slip basis is

= 1 1
v
b
x

l
l
(5.36)

Properties in eqn. 5.35 are based on the liquid. However, the main problem is to determine the mass
velocity, G. In film boiling inside a tube we have a core of liquid surrounded by an annular layer of gas
which is of very low viscosity. No data in the open literature exist for this case and, thus, determining the
circulation rate is a real problem. An alternative estimate of the film coefficient could be made based on
pool boiling correlations.

(e) Maximum Heat Flux for Stable Operation.

In vertical tube thermosyphons there are several limits to operation such as surging, critical heat flux, and
mist flow. The surging instability occurs as T is increased beyond a limit but this surging is dependent
upon the hydraulic layout and is controllable by the physical arrangement. Blumenkrantz and Taborek
(39) discuss the phenomenon. The surging can be controlled by increasing the frictional resistance in the
inlet piping.

Usual recommended design for thermosyphons has the outlet pipe cross-section area equal to the total
cross section area of the tubes. The inlet liquid line is usually smaller and in the range of 25 to 50% of the
outlet pipe area.

As the temperature difference increases, the evaporation rate of a given tube will increase, pass through
a maximum, and then decrease. This was investigated by Lee (40) and confirmed by Palen et.al. (41) and
both presented correlations for this effect. The Palen correlation is preferred due to its simplicity

q
max
= 16066(d
i
/L)
0.35
(P
c
)
0.61
(Pr)
0.25
( 1 Pd) (5.37)

262

5.3.5. Boiling of Mixtures

When a mixture is boiled the heat transfer
coefficient predictions are further complicated by the
effect of the local changes in composition, which in
turn affect physical properties and boiling points.
Some understanding of the factors involved result
from studies on boiling of binary mixtures.

When a binary mixture is boiled, the vapor
generated is richer in the lower boiling point
component and a plot of the boiling points of the
liquid compositions and the dew points of the
resulting vapor compositions is shown in Figure
5.23. As the binary mixture, y
1
is heated the first
vapor bubble forms at the temperature called the
bubble point and that vapor has the composition of
Y
2
as shown in the figure. A plot of these bubble
points versus composition gives the lower (liquid)
curve. In cooling a binary vapor of composition y,
the first drop of condensate forms at the
temperature called the dew point and a plot of dew
points is the vapor curve. Temperatures between
the dew point and bubble point correspond to a
mixture of vapor and liquid each of different
composition but whose sum total composition is y
1
.
These curves in Figure 5.23 are generated under
the special condition of equilibrium between the
vapor and liquid. Boiling is a non-equilibrium
process; however, the formation of bubbles at the
surface depletes the liquid film of the low boiling
component and the remaining liquid has a higher
local boiling point. Thus the effective T from the
surface to the liquid film boiling point is less than the
apparent T and the resulting calculated coefficient
is lower than the actual coefficient. Further, the bulk
liquid temperature is taken as the equilibrium boiling
point of the mixture and ignores any superheat in
the bulk liquid.

(a) Mixture Heat Transfer Coefficients.

Early experiments on mixtures showed, as in Figure
5.24, that the heat fluxes lie between the values for
the pure components. The heat transfer coefficients,
based on the apparent T, are always less than the
pure component coefficients and the minimum
values occur at the concentrations where there is
the greatest separation between the vapor and
liquid lines (42) as illustrated in Figures 5.25 and 5.26.

263

While most mixtures follow the curves as in
Figure 5.23 where one component is more
volatile over the entire concentration range,
there are other systems where one component
is more volatile over only a portion of the
concentration range and less volatile over the
remaining portion. These systems form
azeotropes where the azeotropic composition is
one where the composition of the vapor and
liquid are identical. This azeotropic mixture boils
as though it were a single component liquid.
Examples of such systems (42) are shown in
Figures 5.27 and 5.28. These figures show a
minimum boiling point azeotropic mixture but
maximum boiling point azeotropic mixtures also
exist.

Recent papers by Stephan (42) and by Thomas
(43) give short reviews of some current theories
for heat transfer to boiling mixtures. When we
have multi-component mixtures the theories
become too complicated for design purposes. A
simple empirical approach to calculating mixture
boiling coefficients is based on the paper of
Palen and Small (35) which was later confirmed
as suitable for equipment design in (32) and
recommended by Palen (34). Here a mixture
correction factor, F
m
, is used to modify the
calculated coefficient of the volatile component,
h
nbl
, determined from equation 5.32 so that the
mixture coefficient = h
nbl
F
m
where

F
m
= exp(-0.015 BR) (5.38)

where BR = boiling range, dew point-bubble
point, F. with a lower limit of F
m
= 0. 1. This
relation is shown in Figure 5.29. This equation
is recommended (34) as a reasonable approach
for multicomponent systems. This empirical
equation is based on the boiling range which is
the spread between the vapor and liquid curves
as shown in Figures 5.23, 5.25, and 5.26 and is
close to some of the theoretical methods (42).

(b) Mixture Maximum Heat Flux.

Although studies of maximum heat flux for
mixtures have shown some instances, Figure
5.30, where the maximum flux of a mixture can
be greater than the maximum flux of the
components, (44) there is some disagreement
264

on the cause for some of the higher values and some
question on the adequacy of some theories, see e.g.,
(44,45,46) for details. For design purposes it is
recommended (34) that equation 5.5 be used with the
critical pressure of the mixture based on the molar
average. This will give maximum fluxes lying between
those of the components and will be conservative.












265